Nutanix & Nvidia build unified AI stack for enterprises

Nutanix has expanded its collaboration with Nvidia to launch an integrated AI operating environment, aligning its cloud infrastructure software with Nvidia's latest enterprise AI hardware and software platforms.

The companies are targeting enterprises that are moving generative AI projects from pilot into production and need more predictable infrastructure and management. The combined environment brings together Nutanix's compute, storage, virtualisation and Kubernetes stack with Nvidia AI Enterprise software and upcoming Rubin platform technologies.

Nutanix said its AI-focused operating environment sits on its Acropolis Operating System storage platform and AHV hypervisor. It also draws on the Nutanix Kubernetes Platform, Nutanix Enterprise AI, Nutanix Unified Storage and Nutanix Database products. Canonical's Ubuntu Pro operating system is available for bare metal Kubernetes deployments.

The companies position this as a single operating layer for what they describe as "AI factories". The environment supports deployment of generative AI workloads on Nvidia GPU infrastructure and the Nvidia AI Enterprise software suite, including Nvidia NIM microservices for model inference.

Focus on production

Much of the effort centres on enterprises that have already invested in Nvidia-based infrastructure. Nutanix said the integrated stack is intended to shorten the time between hardware delivery and production use of generative AI models. It also wants customers to run AI workloads under a consistent operational framework rather than stitching together multiple infrastructure tools.

The joint work extends to Nvidia's future hardware architectures. Nutanix plans support for the Nvidia Rubin platform, including Vera Arm-based CPUs and Rubin GPUs, for both bare metal and virtualised environments using AHV and the Nutanix Kubernetes Platform.

Nutanix also intends to support the Nvidia Inference Context Memory Storage Platform. That platform uses the Nvidia BlueField-4 storage processor to handle data for large-scale inference workloads. In addition, Nutanix will integrate Nvidia Spectrum-X Ethernet photonics switch systems to provide an AI-focused data centre interconnect for its AI operating environment.

Token-centric design

Nutanix describes the combined system as designed around the token-based nature of generative AI workloads. Nvidia's GPUs and software handle the intensive token processing and model execution. Nutanix supplies the underlying storage, compute orchestration and virtualisation for those workloads across on-premises and hybrid environments.

The company said its environment aims to give enterprises tighter control over where AI workloads run and how data is managed. This includes support for different deployment models, from bare metal GPU clusters to virtualised clusters and container-based environments.

Nutanix also emphasised its ongoing support for open-source components. The firm stated that it will continue to work with the latest Nvidia open-permissible models and NIM microservices, intending to expose these models through its AI operating environment for a range of customer use cases.

For enterprises, the arrangement offers a pre-integrated route to build generative AI services on infrastructure sourced from both companies. It is also designed to align Nutanix's stack with Nvidia's certified ecosystem around Nvidia AI Enterprise.

Nvidia's view

Nvidia frames the collaboration in the context of what it sees as an AI-driven shift in industrial computing. The company positions the Rubin platform and AI Enterprise software as a foundation for organisations seeking more control over data and deployment.

"The AI-driven industrial revolution requires a foundation that strips away complexity to deliver rapid time to value for enterprise AI," said Justin Boitano, Vice President, Enterprise AI Products, Nvidia. "By integrating the NVIDIA Rubin platform with Nutanix's AI operating environment, organisations gain a sovereign AI foundation to build their own AI factories and turn proprietary data into a decisive competitive advantage."

Nutanix has stated that it will work closely with Nvidia as Rubin and related technologies become generally available. The firm plans additional integrations across future Nvidia hardware and software releases, designed for large-scale inference and training workloads.